The anatomical distribution of brain neurochemistry is thought to be related to serious mental health problems such as suicide and alcoholism. Quantitative studies of the distribution of brain neurochemistry are expensive to perform. To make best use of these valuable data, it is important to use powerful statistical methods. Autoradiography and other brain mapping techniques often generate quantitative measures in a moderate number of brain regions of interest (ROI's), say, 30 to 100. I.e., the number of ROI's is larger than that typically used in ROI analysis of in vivo brain images, but far smaller than the number of voxels in PET or MRI brain images. Special methods are needed to analyze data from this range of numbers of ROI's. Exploratory analysis of this many ROI's is unwieldy. Confirmatory analysis requires more sophisticated methods of adjustment for multiple comparisons than those adequate for a small number of ROI's. However, the """"""""random field"""""""" based methods used in voxel-based analysis of in vivo brain images is also inappropriate. Moreover, brains in neurochemical mapping studies are often matched in pairs, triplets, etc. Such data can be analyzed using the """"""""mixed linear model."""""""" However, this technique is rarely used in neurochemical brain mapping work. Software written by the Principal Investigator makes feasible exploratory analyses of up to 100 or more brain regions. This software will be improved, methods for handling multiple comparisons issues in brain mapping will be developed, and alternatives to testing the simple null hypothesis of """"""""no effect"""""""" will be developed
| Ellis, S P; Underwood, M D; Arango, V et al. (2000) Mixed models and multiple comparisons in analysis of human neurochemical maps. Psychiatry Res 99:111-9 |
